Why Cognitive Restructuring Often Falls Short With Social Anxiety: The Evidence Base

Series: Rethinking Social Anxiety Treatment — Post 1 of 3

If you've been doing this work for a while, you've probably had the experience of watching a client do everything right: completing thought records, identifying cognitive distortions, challenging the evidence, and still not getting better. Not because they aren't trying. Not because you're doing it wrong. Instead, it’s because something about the model isn't quite reaching the problem.

This post is about that ceiling. Not to dismiss cognitive restructuring; it's a well-researched technique with genuine evidence behind it, but to look honestly at where the research shows it running out of room, and what that might tell us about the actual mechanics of social anxiety.

What the Research Actually Shows

Cognitive restructuring (CR) does work for social anxiety. That's worth stating plainly before anything else. Meta-analyses consistently show that CBT, including CR as a component, produces meaningful reductions in social anxiety symptoms compared to waitlist controls (Kindred et al., 2022).

However, here's where it gets more interesting: a review of five meta-analyses found that both exposure and cognitive restructuring reduced social anxiety symptoms in adults, and that combining CR with exposure did not appear superior to exposure alone (Rodebaugh et al., 2004, as cited in de Mooij et al., 2023). That finding has been replicated enough times to be worth sitting with. If CR were doing something irreplaceable in the treatment of social anxiety disorder (SAD), adding it to exposure should consistently improve outcomes. Yet, it doesn't, reliably.

Social anxiety disorder is also associated with some of the lowest remission rates among anxiety-related disorders, hovering around 40% even with first-line CBT treatment (Pozza, 2022). For context, Springer et al. (2018) noted obsessive-compulsive disorder and SAD had the lowest remission rates compared to other anxiety disorders, with generalized anxiety disorder (GAD) having the highest rates, slightly above 60% at follow-up (average 6 months later). That means a substantial portion of clients who receive evidence-based care remain significantly symptomatic, even for GAD. That's a signal that the model may be missing something rather than any clinical error. It also may show that SAD isn’t purely anxiety, the way GAD is.

The Epistemological Problem

Here's what cognitive restructuring assumes: the automatic thoughts driving social anxiety are distorted, inaccurate assessments of threat, probability, or social consequence, and correcting them to “rational reality” reduces distress.

That assumption is reasonable for some presentations. Yet, social anxiety is different in a structurally important way. It's episodic and rooted in habituation. The nervous system has learned, through repeated experience, that social situations carry threat. Those automatic thoughts: I'll embarrass myself, they'll judge me, I'll say something wrong, aren't irrational inventions. They're the nervous system's best prediction based on its history. If anything, those thoughts are inferences.

Telling that system its predictions are distorted isn't just ineffective. In many cases, it may be inaccurate. Much like telling a child “don’t talk to strangers, they may kidnap you” sounds wise, yet an unlikely event that would be called a distortion if life and physical safety were not at stake. However, if the child never talks to someone new, they’ll likely hold that belief in perpetuity until the nervous system has enough disconfirming experience; enough habituation, enough contact with safety, that those predictions are, in a sense, a threat prediction rather than a fact. In social anxiety, the fear of humiliation and body activation are the body's honest read on the situation. After enough repetitions where they’ve felt humiliated, shamed, and like they don’t belong, the body remembers and begins to predict and protect simultaneously. This begins to create a physiological response rather than a purely cognitive one. In a sense, CR asks clients to think their way out of a learned physiological pattern that’s rooted in lived experience.

This deserves attention because prominent cognitive models of SAD, including the Clark and Wells (1995) model and the Rapee and Heimberg (1997) model, both rest on the premise that biased information processing is the core maintenance mechanism. When that premise is only partially true, when the "distortion" is better understood as a conditioned alarm, cognitive challenge alone will consistently undershoot the target. It’s like giving a therapeutic homework assignment to a parent of a little child to go to a mall and have their child go up and talk to strangers, and while that’s happening, look away, oh, and don’t have any “irrational thoughts” about their kids being kidnapped, don’t you know the stats?

The Shame Loop Nobody Talks About

This is where the clinical picture gets more complicated, and where CR creates an inadvertent problem for this particular population.

When a client works hard on thought records week after week and still feels anxious in social situations, they rarely conclude that the model has a ceiling. They conclude they're failing. I must be doing this wrong. Why isn't this working for me? Everyone else seems to get better. Which, over half of people with social anxiety likely will experience this if we only use CR.

Leonov and Laplante (2025) describe shame as an emotion that is highly aversive, unpredictable, and resistant to metacognitive regulation. That last quality is especially important here: trying to think your way out of shame through reason or self-correction tends to backfire. And yet CR, when it doesn't produce the expected results, positions the problem as a cognitive one that the client hasn't solved yet. Taken together, the implicit message is, change your thinking, and the anxiety will follow, and often creates a secondary wound when the anxiety doesn't follow, potentially leading to dropouts and shame.

Research on post-event processing makes this concrete: Cândea and Szentágotai-Tătar (2017) found that state shame, not anxiety severity, not self-evaluation of performance, was the only significant predictor of post-event rumination in socially anxious individuals. In the same study, the authors also noted that self-compassion training appeared easier to implement than cognitive reappraisal, as suggested by lower dropout rates and higher exercise completion accuracy, though they identified this as a preliminary observation worth further investigation.

This all brings up a question: if shame drives the rumination cycle that maintains social anxiety, and the treatment inadvertently activates shame through repeated "failure" to think correctly, are we maybe reinforcing the very mechanism we're trying to interrupt?

The Missing Layer: What the Body Knows First

There's a third problem, one that connects the previous two. Social anxiety is not primarily a cognitive experience. It's also a somatic one.

Ben Shahar (2020) identifies pervasive emotional avoidance, difficulties with emotional differentiation, and high levels of self-criticism as central psychopathological processes in SAD; processes that live in the body and the emotional system, not just in thought content. When a client walks into a social situation, the physiological activation arrives before the thought does. The racing heart, the heat in the face, the chest tightening, these are downstream signals of a nervous system that has already assessed threat. The thought that follows is often an interpretation of the body's alarm, not its cause.

This has direct clinical implications for sequencing. When a client is physiologically dysregulated, outside their window of tolerance, cognitive flexibility is severely limited (Ogden et al., 2006). Asking someone to examine the accuracy of their thoughts while their nervous system is in a threat state is asking the wrong system to do work it can't do in that moment. Nervous system regulation needs to come first. Cognitive work becomes available after.

Standard CR protocol doesn't account for this sequencing. It treats the thought as the entry point, bypassing the somatic and emotional experience that generated it. In hindsight, this may actually be reinforcing experiential avoidance.

What gets missed, in these cases, isn't the cognitive content of the anxiety. It's the emotional experience underneath it: the shame, the sense of not belonging, the feeling of fundamental exposure. These aren't distortions to correct. They're experiences to make contact with.

This distinction is important, especially for clients who have difficulty identifying and naming internal states, those who experience the alarm without being able to clearly locate it, describe it, or separate what they feel from what they think. For these clients, a cognitive intervention isn't just limited in what it can reach. It's addressing the wrong level of the system entirely.

What This Means Clinically

None of this means CR has no place in social anxiety treatment. Cognitive work has real value, particularly later in the process, once the nervous system has enough regulatory capacity to examine thought patterns with some flexibility. The sequencing problem, not the technique itself, is often what limits outcomes.

Yet if a client isn't responding to CR, the most useful question probably isn't, "How do I help them challenge these thoughts more effectively?" It may be: what is this person's nervous system actually needing, and am I starting at the right level?

The research suggests that for a meaningful proportion of clients with social anxiety, the ceiling of CR isn't a skill deficit on their part or a delivery problem on ours. It's a model problem, a case of the right tool aimed at the wrong level of the system. Cognitive restructuring was designed to address distorted thinking. Social anxiety, at its core, may be something else: an emotional and somatic experience of shame, threat, and disconnection that calls for a different kind of contact before we address thinking.

That's what the next post in this series takes up; why the ACT model may address shame more directly than CBT does, and what that difference looks like mechanistically and in practice.

References

Ben Shahar, B. (2020). New developments in emotion-focused therapy for social anxiety disorder. Journal of Clinical Medicine, 9(9), Article 2918. https://doi.org/10.3390/jcm9092918

Cândea, D.-M., & Szentágotai-Tătar, A. (2017). Shame as a predictor of post-event rumination in social anxiety. Cognition and Emotion, 31(8), 1684–1691. https://doi.org/10.1080/02699931.2016.1243518

Clark, D. M., & Wells, A. (1995). A cognitive model of social phobia. In R. G. Heimberg, M. R. Liebowitz, D. A. Hope, & F. R. Schneier (Eds.), Social phobia: Diagnosis, assessment, and treatment (pp. 69–93). Guilford Press.

de Mooij, B., Fekkes, M., Miers, A. C., van den Akker, A. L., Scholte, R. H., & Overbeek, G. (2023). What works in preventing emerging social anxiety: Exposure, cognitive restructuring, or a combination? Journal of Child and Family Studies, 32, 498–515. https://doi.org/10.1007/s10826-023-02536-w

Kindred, R., Bates, G. W., & McBride, N. L. (2022). Long-term outcomes of cognitive behavioural therapy for social anxiety disorder: A meta-analysis of randomised controlled trials. Behaviour Research and Therapy, 150, Article 104017. https://doi.org/10.1016/j.brat.2022.104017

Leonov, A., & Laplante, J. P. (2025). Shame or what makes irrational social anxiety rational. Healthcare, 13(22), Article 2891. https://doi.org/10.3390/healthcare13222891

Ogden, P., Minton, K., & Pain, C. (2006). Trauma and the body: A sensorimotor approach to psychotherapy. Norton.

Pozza, A. (2022). Efficacy of psychological and psychiatric treatments and potential predictors in social anxiety disorder and obsessive-compulsive disorder [Editorial]. Frontiers in Psychiatry, 12, Article 833131. https://doi.org/10.3389/fpsyt.2021.833131

Rapee, R. M., & Heimberg, R. G. (1997). A cognitive-behavioral model of anxiety in social phobia. Behaviour Research and Therapy, 35(8), 741–756. https://doi.org/10.1016/S0005-7967(97)00022-3

Rodebaugh, T. L., Holaway, R. M., & Heimberg, R. G. (2004). The treatment of social anxiety disorder. Clinical Psychology Review, 24(8), 883–908. https://doi.org/10.1016/j.cpr.2004.07.007

Springer, K. S., Levy, H. C., & Tolin, D. F. (2018). Remission in CBT for adult anxiety disorders: A meta-analysis. Clinical Psychology Review, 61, 1-8. https://doi.org/10.1016/j.cpr.2018.03.002